Increasing the accuracy of a statement by analyzing the relationships between entities in a knowledge graph

1. A method of improving the accuracy of a natural language statement comprising: receiving the natural language statement from a user in computer-readable form, by executing first instructions in a computer system; extracting from the natural language statement a plurality of entities and at least one stated relationship using a natural language processing engine including a machine learning model, wherein the stated relationship is between at least a first one of the entities and a second one of the entities and is based on a target phrase in the statement, by executing instructions in the computer system; determining that the natural language statement is inaccurate, by executing third instructions in the computer system, based, at least in part, on a first numeric accuracy score representing an overall accuracy of the natural language statement; identifying a predominant relationship between the first and second entities using a knowledge base wherein the predominant relationship is different from the stated relationship, by executing fourth instructions in the computer system; generating a candidate phrase representing the predominant relationship using natural language processing, by executing fifth instructions in the computer system; computing a second numeric accuracy score representing an overall accuracy of the candidate statement, by executing instructions in the computer system; establishing that the second numeric accuracy score is greater than a sum of the first numeric accuracy score and a predetermined tuning value, by executing instructions in the computer system; and presenting the candidate phrase to the user as a suggested replacement for the target phrase in the natural language statement, by executing instructions in the computer system.

2. The method of claim 1 wherein said identifying includes: constructing a knowledge graph using the knowledge base wherein the knowledge graph includes a plurality of vertices representing the entities in the natural language statement and a plurality of edges representing relationships between the entities; and the predominant relationship is based on a statistically significant number of occurrences of a candidate relationship along a particular one of the edges of the knowledge graph, where statistically significant corresponds to more colocations of the predominant relationship than any other relationships.

3. The method of claim 1 wherein: said identifying identifies multiple predominant relationships between the first and second entities; said generating generates multiple candidate phrases representing the multiple predominant relationships; and said presenting presents the multiple candidate phrases to the user as suggested replacements for the target phrase in the natural language statement.

4. The method of claim 1 wherein said first numeric accuracy score is computed using a yes/no pipeline which uses cognitive analysis of the natural language statement based on information contained in the knowledge base.

5. The method of claim 4 wherein: the natural language statement is determined to be inaccurate when the numeric accuracy score is below a programmable threshold.

6. The method of claim 1 wherein the natural language statement is a query for submission to a deep question answering system, and further comprising: automatically replacing the target phrase in the natural language statement with the candidate phrase to form a modified statement; and submitting the modified statement to the deep question answering system.

7. A computer system comprising: one or more processors which process program instructions; a memory device connected to said one or more processors; and program instructions residing in said memory device for improving the accuracy of a natural language statement by: receiving the natural language statement from a user in computer-readable form, extracting from the natural language statement a plurality of entities and at least one stated relationship using a natural language processing engine including a machine learning model, wherein the stated relationship is between at least a first one of the entities and a second one of the entities and is based on a target phrase in the statement, determining that the natural language statement is inaccurate, based, at least in part, on a first numeric accuracy score representing an overall accuracy of the natural language statement, identifying a predominant relationship between the first and second entities using a knowledge base wherein the predominant relationship is different from the stated relationship, generating a candidate phrase representing the predominant relationship using natural language processing, computing a second numeric accuracy score representing an overall accuracy of the candidate phrase, establishing that the second numeric accuracy score is greater than a sum of the first numeric accuracy score and a predetermined tuning value, and presenting the candidate phrase to the user as a suggested replacement for the target phrase in the natural language statement.

8. The computer system of claim 7 wherein said program instructions further construct a knowledge graph using the knowledge base, the knowledge graph including a plurality of vertices representing the entities in the natural language statement and a plurality of edges representing relationships between the entities, and the predominant relationship is based on a statistically significant number of occurrences of a candidate relationship along a particular one of the edges of the knowledge graph, where statistically significant corresponds to more colocations of the predominant relationship than any other relationships.

9. The computer system of claim 7 wherein said program instructions identify multiple predominant relationships between the first and second entities, generate multiple candidate phrases representing the multiple predominant relationships, and present the multiple candidate phrases to the user as suggested replacements for the target phrase in the natural language statement.

10. The computer system of claim 7 wherein said first numeric accuracy score is computed using a yes/no pipeline which uses cognitive analysis of the natural language statement based on information contained in the knowledge base.

11. The computer system of claim 10 wherein the natural language statement is determined to be inaccurate when the numeric accuracy score is below a programmable threshold.

12. The computer system of claim 11 wherein said program instructions further generate a candidate statement by substituting the candidate phrase for the target phrase in the natural language statement.

13. The computer system of claim 7 wherein the natural language statement is a query for submission to a deep question answering system, and said program instructions further automatically replace the target phrase in the natural language statement with the candidate phrase to form a modified statement, and submit the modified statement to the deep question answering system.

14. A computer program product comprising: a computer readable storage medium; and program instructions residing in said storage medium for improving the accuracy of a natural language statement by: receiving the natural language statement from a user in computer-readable form, extracting from the natural language statement a plurality of entities and at least one stated relationship using a natural language processing engine including a machine learning model, wherein the stated relationship is between at least a first one of the entities and a second one of the entities and is based on a target phrase in the statement, determining that the natural language statement is inaccurate, based, at least in part, on a first numeric accuracy score representing an overall accuracy of the natural language statement, identifying a predominant relationship between the first and second entities using a knowledge base wherein the predominant relationship is different from the stated relationship, generating a candidate phrase representing the predominant relationship using natural language processing, computing a second numeric accuracy score representing an overall accuracy of the candidate phrase, establishing that the second numeric accuracy score is greater than a sum of the first numeric accuracy score and a predetermined tuning value, and presenting the candidate phrase to the user as a suggested replacement for the target phrase in the natural language statement.

15. The computer program product of claim 14 wherein said program instructions further construct a knowledge graph using the knowledge base, the knowledge graph including a plurality of vertices representing the entities in the natural language statement and a plurality of edges representing relationships between the entities, and the predominant relationship is based on a statistically significant number of occurrences of a candidate relationship along a particular one of the edges of the knowledge graph.

16. The computer program product of claim 14 wherein said program instructions identify multiple predominant relationships between the first and second entities, generate multiple candidate phrases representing the multiple predominant relationships, and present the multiple candidate phrases to the user as suggested replacements for the target phrase in the natural language statement.

17. The computer program product of claim 14 wherein said first numeric accuracy score is computed using a yes/no pipeline which uses cognitive analysis of the natural language statement based on information contained in the knowledge base.

18. The computer program product of claim 17 wherein the natural language statement is determined to be inaccurate when the numeric accuracy score is below a programmable threshold.

19. The computer program product of claim 18 wherein said program instructions further generate a candidate statement by substituting the candidate phrase for the target phrase in the natural language statement.